MODIS imagery of turbid plumes in San Diego coastal waters during rainstorm events

Remote Sensing of Environment 114 (2010) 332–344

Contents lists available at ScienceDirect

Remote Sensing of Environment j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / r s e

MODIS imagery of turbid plumes in San Diego coastal waters during rainstorm events Florence Lahet, Dariusz Stramski ⁎ Marine Physical Laboratory, Scripps Institution of Oceanography, University of California San Diego, 9500 Gilman Drive, La Jolla, CA 92093-0238, USA

a r t i c l e

i n f o

Article history: Received 6 November 2008 Received in revised form 13 August 2009 Accepted 20 September 2009 Keywords: Ocean color MODIS Coastal ocean Turbid plumes Rainstorm events Southern California

a b s t r a c t Data of normalized water-leaving radiance, nLw, obtained from the Moderate Resolution Imaging Spectroradiometer (MODIS) on the Aqua satellite at spatial resolution of 250 m (band 1 centered at 645 nm) and 500 m (band 4 at 555 nm) are used to study turbid plumes in coastal waters of southern California during rainstorm events in winter of 2004–2005. Our study area includes San Diego coastal waters, which extend approximately 25 km offshore between Point Loma and 10 km south of the US–Mexican border. These waters are influenced by terrigenous input of particulate and dissolved materials from San Diego and Tijuana watersheds and non-point sources along the shore. Optimum threshold values of satellite-derived normalized water-leaving radiances at both wavebands were established for distinguishing the plume from ambient ocean waters. These threshold values were determined by searching for a maximum correlation between the estimates of satellite-derived plume area calculated using a broad range of nLw values and the environmental variables characterizing rainfall, river discharge, wind, and tides. A correlation analysis involving the amount of precipitated water accumulated during a storm event over the San Diego and Tijuana watersheds was selected as the basis for final determinations of the optimum threshold nLwthr and subsequent calculations of the plume area. By applying this method to a sequence of MODIS imagery, we demonstrate the spatial extent and evolution of the plume during rainstorm events under various conditions of precipitation, river discharge, wind forcing, and coastal currents. © 2009 Elsevier Inc. All rights reserved.

1. Introduction Ocean color remote sensing offers an attractive approach to distinguish turbid plumes produced by stormwater discharge in the coastal ocean from ambient marine waters. Optical sensors deployed on aircrafts or satellites provide a means to detect plumes over extended spatial and temporal scales that cannot be adequately addressed with traditional analysis of discrete water samples. Plumes are influenced by various factors such as the nature and magnitude of runoff discharged from rivers and other sources, wind, currents, tides, and the buoyancy of water (Stumpf et al., 1993; Garvine, 1995; Chao, 1998; Warrick et al., 2004b; Ahn et al., 2005). Several studies of stormwater plumes in the Southern California Bight showed the potential usefulness of ocean color satellite data for water quality assessment and coastal management (e.g., Ahn et al., 2005; Warrick et al., 2007; Nezlin et al., 2008). Using satellite observations from the Sea-viewing Wide Field-ofview Sensor (SeaWiFS), Nezlin and DiGiacomo (2005) analyzed the relationship between the amount of precipitated rainwater and plume characteristics over the San Pedro Shelf, which is adjacent to the coastal watershed within the Los Angeles metropolitan area. By assessing maximum correlation between the plume size and precipitated rainwater, they found that satellite-derived normalized water-leaving

⁎ Corresponding author. Marine Physical Laboratory, Scripps Institution of Oceanography, University of California San Diego, La Jolla, 9500 Gilman Dr, CA 92093-0238, USA. E-mail address: [email protected] (D. Stramski). 0034-4257/$ – see front matter © 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.rse.2009.09.017

radiances, nLw, above the threshold value of 1.3 mW cm− 2 μm− 1 sr− 1 at 555 nm allowed discrimination of the runoff plume from ambient ocean waters. Nezlin et al. (2005) compared the relationships between plume size derived from SeaWiFS and rainstorm data in different coastal areas of southern California. They showed that the primary factors controlling the relationships include watershed land-use characteristics, watershed size, and land topography. The objective of this study is to analyze the satellite-derived plume area in relation to environmental parameters during rainstorm events in the San Diego region of southern California. We use ocean color imagery from the Moderate Resolution Imaging Spectroradiometer (MODIS) flown aboard the Aqua spacecraft. Two high spatial resolution bands, 1 and 4, within the visible spectral region are used. The MODIS band 1 with a spatial resolution of 250 m has a spectral range of 620–670 nm with a center wavelength of 645 nm. The band 4 with a resolution of 500 m has a spectral range of 545–565 nm centered at 555 nm. Recent studies demonstrated the potential of these bands to monitor water quality in estuarine and coastal waters (Hu et al., 2004; Chen et al., 2007; Shutler et al., 2007). We examine the time period from December 28, 2004 through March 7, 2005, which represents the third heaviest rainfall season in southern California since records began in 1850 (NOAA National Weather Service, 2007). During that period, repeated and at times long periods of intense rain and runoff had a significant effect on coastal water quality in the San Diego region. We defined several storm events using rainfall data and analyzed eighteen MODIS images. We discuss the determinations of threshold values for satellite-derived water-leaving

F. Lahet, D. Stramski / Remote Sensing of Environment 114 (2010) 332–344

radiances, nLw(645) at 645 nm and nLw(555) at 555 nm, which are used to define a plume. These threshold determinations are based on examining the correlation between the plume area and environmental parameters (rainfall, Tijuana River discharge, wind, tides). Using a sequence of MODIS images, we illustrate the spatial extent and evolution of the plume during selected rainstorm events. 2. Characterization of the study area The study area extends along the San Diego coastline from approximately Point Loma in the north to 10 km south of the US–Mexican border, and is defined by latitudes of about 32°26′N and 32°41′N and longitudes 117°04′W and 117°22′W (Fig. 1). This area includes the coastal ocean waters extending about 25 km offshore. The enclosed waters of San Diego Bay are not considered for plume determinations, as our interest is focused just on coastal ocean waters adjacent directly to the open ocean. The San Diego Bay is connected with the ocean near Point Loma. The environmental problems in the study area occur due to the large population of the San Diego-Tijuana metropolitan area and the concentrated commercial, naval, and recreational activities (Schiff et al., 2000). There are multiple sources of particles, organic substances, nutrients, and contaminants that discharge into the coastal ocean in this region (Tran et al., 1997; Zeng & Vista, 1997; Zeng et al., 1997). Significant degradation of coastal water quality is caused by stormwater runoff during episodic rainstorm events, mainly in the winter season (e.g., Characklis & Wiesner, 1997; Davis et al., 2001). In particular, the Tijuana River discharges into the ocean just north of the US–Mexican border. After heavy rains, the Tijuana River carries runoff from the city of Tijuana and from sewage that overflows from the International Wastewater Treatment Plant in Tijuana. Plumes from the river can travel north along

333

the San Diego coast at least as far as Point Loma, or south along the Mexican coast of northern Baja California. The storm drain system also contributes to pollution of coastal waters. Consequently, turbid plumes in coastal waters, especially during storms, represent an environmental hazard due to associated pollutants. The beaches of Imperial Beach north of the Tijuana River mouth were closed during 83 days in 2005. Our study region is characterized by an arid warm Mediterranean climate with 85% of precipitation occurring during a rainy season from November through March. The average annual precipitation is less than 300 mm. Rainfall is highly variable from year to year and from month to month. Droughts are typical. Floods occur at times, although they are generally short lasting. Climate in the region can vary considerably over short geographical distances. For example, the western slope of the Peninsular Range (the prominent topographic feature of the region) receives about 1 m and the foothills west of the Peninsular Range about 400–500 mm of annual rainfall (Isla & Lee, 2006). There are eleven hydrologic units in the San Diego region, which are associated with river/stream systems that discharge into the Pacific Ocean (Fig. 2). We have considered the following hydrologic units (from south to north): Tijuana (TIJ), Otay (OT), Sweetwater (SW), Pueblo San Diego (PUE), San Diego (SD), Penasquitos (PEN), and San Dieguito (SDo) (see also Nezlin & Stein, 2005). Whereas the southernmost units (TIJ and OT) are adjacent to the coastal waters of direct interest to our study, the remaining units may also affect the area through the outflow from San Diego Bay or predominant southward transport of coastal waters from the north. The rivers of the region are small but have generally high sediment yields. For example, Tijuana, Sweetwater, and San Diego rivers are characterized by a mean annual flow of 28.9× 106, 7.42 × 106, and 13.7× 106 m3 yr− 1 and a mean annual suspended sediment flux of 0.206 × 106, 0.0043 × 106, and 0.010 × 106 ton yr− 1, respectively (Inman & Jenkins, 1999).

Fig. 1. Map of southern California coastal waters showing the study area (black box). The map specifies the location of Mission Bay (MB), San Diego River mouth (SDR), Point Loma (PL), Coronado (C), San Diego Bay (SDBa), Tijuana River (TR), the US–Mexican border (B), Los Buenos Creek mouth and Punta Los Buenos (LBC and PLB). The bathymetry is shown. The four sections, S1, S2, S3, and S4, for which the offshore extent of plume was calculated (Table 3 and relevant text in Section 4.3) are also shown as dashed lines.

334

F. Lahet, D. Stramski / Remote Sensing of Environment 114 (2010) 332–344

Fig. 2. Thirty-seven stations (solid circles) located within the US part of the Tijuana watershed and the San Diego watershed (Otay, Sweetwater, Pueblo San Diego, San Diego, Penasquitos, and San Dieguito hydrologic units), which provided data for calculating the rainfall parameters. The map was obtained from the website of the San Diego watershed network (http://map.sdsu.edu/group2001/group3/index.htm). Stations within the Mexican part of the Tijuana hydrologic unit are not shown.

Winds in the San Diego region are generally weak. Wind speeds in excess of 8 m s− 1 are infrequent (Lentz & Winant, 1986). A sea breeze, which is common from the late morning through the evening, produces a westerly flow at an average speed of about 4–5 m s− 1. At times, mainly in the fall and winter, dry and windy weather conditions known as Santa Ana form in southern California (e.g., Raphael, 2003). Ocean currents in the study area are related to different forcing including barotropic tides, internal waves, winds, as well as larger-scale flows such as the southward flowing California Current system (Bray et al., 1999; Chadwick & Largier, 1999). In addition, current patterns are related to regional bathymetry and coastline shape (Kim et al., 2007; Jan Svejkovsky, personal communication). The prominent bathymetric features include the relatively steep drop-off paralleling the outside edge of the Point Loma kelp bed, the shoreward reaching deep-water basin outside Coronado Shores, the alluvial fan of the Tijuana River, and the gradually sloping bottom along a relatively straight coastline from Coronado to Punta Los Buenos in Mexico (Fig. 1). The most frequent flow regime (occurring 60–70% of the time) is dominated by moderate southward currents. The flow immediately south of Point Loma tends to round the headland toward the east. Near the beach, most of this water again veers southward until deflected offshore by the shallows of the Tijuana River alluvial fan. This alluvial fan tends to sometimes create a cyclonic eddy offshore from the Tijuana Estuary in the US and Playas de Tijuana in Mexico. Northward flow regime is relatively uncommon (20% of the time) and generally lasts 1 to 2 days. The most significant feature during northward flow is related to the influence of the Tijuana River alluvial fan. As the near-shore waters reach the shallows from the south, they become deflected to the northwest (i.e., offshore). The tidal range in San Diego Bay is ∼1.7 m from mean lower-low to mean higher-high water, with extreme range up to 3 m (Chadwick & Largier, 1999). Cross-shelf and alongshore tidal currents on the Southern California shelf are poorly correlated with the tidal elevation (Winant & Bratkovich, 1981).

3. Data and methods 3.1. Satellite data MODIS-Aqua Level 1A data were obtained from NASA Goddard Space Flight Center and processed to Level 2 format using the NASA's SeaWiFS Data Analysis System (SeaDAS version 5.1.6) software. The normalized water-leaving radiances nLw(645) (i.e., band 1 at 645 nm, 250 m spatial resolution) and nLw(555) (i.e., band 4 at 555 nm, 500 m spatial resolution) were calculated. The atmospheric correction was based on an aerosol model utilizing the shortest infrared wavelength at 1240 nm and the longest infrared wavelength at 2130 nm (Wang & Shi, 2005; Wang, 2007). Eighteen MODIS images for the period from 12/30/2004 through 3/7/2005 were processed for the study area and projected using a Lambert azimuthal equal area projection. During that time period several significant rainstorm events occurred. 3.2. Precipitation data Daily rainfall data from 44 meteorological stations located in seven hydrologic units (Fig. 2) were obtained from NOAA National Climatic Data Center Climate Data Online, the San Diego County rainfall monitoring network, and the Mexican “Comisión Nacional del Agua”. A mean daily precipitation, Ph̄ , was first calculated for each hydrologic unit, h. ̄ , for We then calculated an area-weighted mean daily precipitation, PWS composite watersheds, WS, from:

=

n — — n P WS = ∑ Ah P h ∑ Ah h=1

h=1

ð1Þ

where Ah is the area of the h-th hydrologic unit and n is the number of hydrologic units included in the composite watershed (Nezlin & Stein, 2005). Seven composite watersheds were considered with an

F. Lahet, D. Stramski / Remote Sensing of Environment 114 (2010) 332–344

increasing spatial extent from south to north. The first composite watershed, WS1, represents just the Tijuana hydrologic unit, the second composite watershed, WS2, includes Tijuana and Otay units, the third WS3 includes Tijuana, Otay, and Sweetwater units, etc. The final composite watershed, WS7, includes seven hydrologic units extending from the Tijuana unit in the south to the San Dieguito unit in the north (see Fig. 2). The Tijuana unit is the largest unit with most of its area in Mexico. The large part of the Mexican area has, however, relatively few rainfall stations. Therefore, in our calculations of rainfall parameters, we tested two versions of the Tijuana unit, with and without the Mexican part. For defining the beginning date and the end date of an individual rainstorm event, we followed the criteria proposed by Nezlin and DiGiacomo (2005). According to these criteria, the rainstorm begins on a day when the accumulated precipitation during the 7-day period preceding that day exceeds 2.5 mm. Under this threshold value, no significant rainfall effect on water quality in near-shore waters was observed in the Los Angeles region (Ackerman & Weisberg, 2003). The values of area-weighted mean daily precipitation were considered significant when greater than 0.25 mm. The lower values were ignored. ̄ < 0.25 mm and an The end of a storm was defined by a day with PWS accumulated precipitation during the 5-day period following that day less than 2.5 mm.

335

̄ Using these criteria, we analyzed daily precipitation data of PWS during the time period December 28, 2004–March 7, 2005 and we identified three storm events (Fig. 3a and Table 1). The first event between December 28, 2004 and January 12, 2005 was characterized by ̄ > 20 mm, including one day with 3 days of heavy rains with PWS ̄ = 45 mm. The second event between January 25 and January 29, PWS 2005, is a rainy episode lasting 5 days with daily precipitation in the range from about 0.3 to 7 mm. The third event lasted from February 6 through March 6, 2005 and consisted of a series of heavy rain events separated by a few days of weak or no precipitation. The dates indicating the beginning and the end of each event as well as the dates when MODIS imagery is available for the analysis of plumes during these events are given in Table 1. We also calculated the accumulated amount of precipitated water, Vt, for the day t of the plume analysis with MODIS imagery. The value of Vt represents the precipitated water accumulated during the period of a given storm and was calculated from: — — 2— 3— 4— Vt = P WS;t−1 + k P WS;t−2 + k P WS;t−3 + k P WS;t−4 + k P WS;t−5 + … ð2Þ where k is the coefficient characterizing the persistence of the precipitated water within the plume, and t − 1, t − 2, t − 3, etc. are the

Fig. 3. Variations of selected environmental factors during the time period from December 28, 2004 through March 7, 2005. This time period is indicated at the top axis. (a) An areaweighted mean daily precipitation (solid line) and accumulated amount of precipitated water, Vt (dashed line is for k = 0.5 and dashed–dotted line is for k = 0.7). (b) Mean daily discharge of Tijuana River, Q. (c) Mean daily wind vectors (W is wind speed). (d) Tidal phase, TP. The consecutive numbers from 1 through 18 at the bottom horizontal axis correspond to MODIS image number (see Table 1). The arrows at the top axis designate the dates of the 6 MODIS images shown in Figs. 8 and 9.

336

F. Lahet, D. Stramski / Remote Sensing of Environment 114 (2010) 332–344

Table 1 Occurrence of rainstorm events determined from the analysis of an area-weighted ̄ . mean daily precipitation, PWS Event

Beginning and end date

MODIS imagery

1

12/28/2004 1/12/2005 1/25/2005 1/29/2005 2/6/2005

12/30/2004 (1); 1/1/2005 (2); 1/12/2005 (3); 1/19/2005 (4)

2 3

3/6/2005

1/29/2005 (5); 1/31/2005 (6); 2/2/2005 (7); 2/3/2005 (8) 2/9/2005 (9); 2/13/2005 (10); 2/16/2005 (11); 2/18/2005 (12); 2/23/2005 (13); 2/24/2005 (14); 2/25/2005 (15) 2/27/2005 (16); 3/2/2005 (17); 3/7/2005 (18)

Shown are the beginning and end dates of the events as well as the dates of MODIS-Aqua images available for the analysis of plumes associated with the events during the time period December 28, 2004–March 7, 2005. The MODIS image numbers are shown in parenthesis.

indices for consecutive days during the period preceding the day t (Nezlin & DiGiacomo, 2005). The highest order term in the sum of Eq. (2) corresponds to the day before the beginning of a given storm event. Vt was calculated for k varying between 0.05 and 0.95 with a step of 0.05.

3.3. Other environmental data Data of Tijuana River discharge were obtained from the San Diego Coastal Ocean Observing System (SDCOOS), Scripps Institution of Oceanography. The flow rate, Q, of Tijuana River measured every 15 min by a river gauge located at the US–Mexican border (32°32′53″ N, 117°03′1.8″W) was used to determine daily average values of flow rate (Fig. 3b). Wind speed and direction were measured at San Diego Lindbergh Field (32°44′N, 117°10′W). This airport is located close to the northern part of the study area and collects daily meteorological data available from NOAA National Climatic Data Center. Wind speed was quantified in terms of daily “resultant wind speed”, i.e., the vector sum of the wind speeds divided by the number of observations. The wind vectors calculated from these data are shown in Fig. 3c. Wind stress was also estimated from a simple algorithm of Large and Pond (1981). The meridional (alongshore) and zonal (onshore) wind stress, τNS and τEW respectively, were determined for each day of the investigated time period. Accumulated values of Tijuana River flow, wind speed, and alongshore and onshore wind stress were calculated similarly to the accumulated rainfall Vt (see Eq. (2)). Analogous to rainfall, these calculations were made to take into account the potential integrated effect of a given environmental variable on the plume over several preceding days during the “life time” of the storm event. In our notation, the subscript t indicates that the variable in question represents the accumulated value, for example Qt is the accumulated flow of Tijuana River. The tidal variations were estimated from the WXTide32 software package (Free Software Foundation), which predicts tides using the algorithm of NOAA National Ocean Service. The daily tides were computed in the northern part of the San Diego Bay (32°42′48″N, 117°10′24″W). This location is close to the northern boundary of the study area. We restrict our considerations to the tidal phase (TP), which is defined as the time from the last high slack water to the time of MODIS image acquisition (Dinnel et al., 1990). This parameter was calculated for each MODIS image within the investigated period (Fig. 3d). In addition, we obtained data of surface currents within the study area from SDCOOS. The current measurements were made using four HF radars located in the region (Roughan et al., 2005; Kim et al., 2007). Each radar site reports radial vectors on an hourly basis, which are combined to yield a total vector solution for producing maps of surface currents on a daily basis.

3.4. Plume area calculation The calculation of plume area, PA, includes two main components. First, a threshold value for the satellite-derived normalized waterleaving radiance, nLwthr, is determined to identify the boundary of the plume, and second, the area of satellite pixels within the plume is calculated. These determinations were made with the MODIS imagery of nLw(645) and nLw(555). The plume area obtained from nLw(645) is denoted by PA645 and that obtained from nLw(555) by PA555. In these determinations, nLwthr was varied between 0.01 and 0.05 with a step of 0.01 mW cm− 2 μm− 1 sr− 1 for both MODIS bands and then between 0.05 and 2 (band 1) and between 0.05 and 3 (band 4) with a step of 0.05 mW cm− 2 μm− 1 sr− 1. We estimated the optimum threshold values of nLwthr at 645 nm and 555 nm for distinguishing the plume from ambient ocean waters by searching for the maximum correlation between the plume area and the accumulated rainfall, Vt. These optimum levels of nLwthr are assumed to yield the best estimates of plume area. Although Vt is the primary environmental factor providing a basis for determining the optimum levels of nLwthr, we also present similar calculations based on other environmental factors (Section 4.2 below). The area of the plume within boundaries determined by nLwthr was calculated using a method based on the multiplication of the matrix of pixel areas by a mask, masknLw. The masknLw value for a given pixel equals to 1 when nLw ≥nLwthr, and 0 when nLw
F. Lahet, D. Stramski / Remote Sensing of Environment 114 (2010) 332–344

Fig. 4. The coefficient of determination, r2, of the linear relationship between the plume area, PA, and the accumulated amount of precipitated water, Vt, as a function of the threshold value of satellite-derived normalized water-leaving radiance, nLwthr. The correlation curves are plotted for various values of parameter k. Panel (a) is for the MODIS band 1 centered at 645 nm and panel (b) for the band 4 at 555 nm.

differences. For example, Nezlin et al. (2005) examined a larger coastal region during a longer period of time (October 6, 1997–June 26, 2003) compared with our study. The relatively short period of time covered in our study represents a season with very heavy rainfall. This is expected to result in more rapid saturation of soils with water and discharge of stormwater to coastal ocean, which can reduce k. In addition, highly turbid plumes during heavy rainfall produce enhanced water-leaving radiance. This may explain the higher value of optimum nLwthr(555) in our study. Nezlin et al. (2005) also showed that land-use characteristics, size, and elevation of the drainage area are important factors affecting coastal plumes. Based on effective land use, impervious cover of the watersheds of Orange County and San Diego region is <10%, so the amount of water infiltrating soils is expected to be relatively high and the runoff to the ocean relatively slow. Such conditions are expected to favor relatively high values of k. Fig. 5 shows the actual data of plume areas, PA645 and PA555, plotted versus Vt for our study area. The plume areas were calculated using the optimum values of nLwthr and k, as determined from Fig. 4. For both MODIS bands, the data points are spread over the observed range of variability. PA645 varies between nearly zero and about 80 km2 and PA555 between a few km2 and 90 km2. The intercept of the relationships is positive but small, suggesting that rainfall was a primary source of the plumes. Some images were not obtained under perfectly clear skies, which contribute to the scatter in the data points. Some outliers may also originate from the presence of surface water impoundments in the

337

Fig. 5. The plume area, PA, as a function of the accumulated amount of precipitated water, Vt, calculated using the optimum threshold values of normalized water-leaving radiance nLwthr and parameter k given in Table 2. Panel (a) shows the relationship for the MODIS band 1 centered at 645 nm, and panel (b) for the band 4 at 555 nm. The data encircled were collected during large Tijuana river discharge exceeding 10 m3 s− 1.

watersheds. The water levels in the reservoirs were very low prior to the rainstorms of 2004–2005 and held back major volumes of precipitated water during the rainstorms. During these rainstorms, there was also significant resuspension through wave action, which cannot be distinguished from older runoff in MODIS data. Consequently, the plumes may sometimes appear larger than their true runoff components. The consideration of rainfall data representing the composite watershed WS7 (including the hydrological units from Tijuana to San Dieguito) appears to be a reasonable approach for our analysis. The highest values of the coefficient of determination, r2 = 0.75–0.77, between the plume area and Vt are obtained for this composite watershed. During rainstorm events the plume is generally affected by multiple terrestrial sources of freshwater discharge into the ocean, which can extend over significant distances along the coastline. Thus, the hydrologic units located north of our study area may have influence on the plume within the study area, especially as the transport of surface waters in this region is predominantly southward. The consideration of the Tijuana unit alone yields the lowest values of r2. Nevertheless, these values are relatively high (0.65–0.68), which reflects a major impact of the Tijuana unit on the study area. The addition of data from the Mexican part of the Tijuana unit does not improve the correlation significantly compared with the consideration of the US part only.

338

F. Lahet, D. Stramski / Remote Sensing of Environment 114 (2010) 332–344

4.2. Relationships between the plume area and various environmental parameters Table 2 provides a summary of relationships between the plume area, PA, and various environmental parameters for both MODIS ̄ bands. In addition to the rainfall data, PWS and Vt, the parameters characterizing the Tijuana River flow, wind conditions, and tides are considered in this analysis. A multivariate analysis of more than two variables at a time was not conducted because the number of available data is too small. The highest values of r2 (>0.75) are found for the linear relationships between PA and the accumulated rainfall, Vt, the accumulated wind speed, Wt, and the accumulated zonal wind stress, τEW,t. Weak correlation is observed between PA and the tidal phase and no significant correlation for the meridional wind stress. Among ̄ , the daily values of environmental parameters, only the rainfall, PWS and Tijuana River flow, Q, show significant correlation with PA (r2 = 0.53–0.63). The Tijuana River flow is the only parameter for which the accumulated values Qt do not correlate better with PA than the mean daily values Q. In addition, our data set shows no correlation (r2 is only 0.0039) between the mean daily rainfall within the Tijuana hydrologic unit and the mean daily discharge of Tijuana River. A possible explanation is the diversion, prior to reaching the Tijuana Estuary, of up to 45,500 m3 per day of the Tijuana River flow into the International Wastewater Treatment Plant. This flow is then treated and released through an offshore outfall, so there is no Tijuana River outflow into the ocean until the flow volume exceeds about 45,500 m3 per day. Generally, this corresponds to more than about 2.5 mm of rainfall (Jan Svejkovsky, personal communication). The variation of r2 as a function of nLwthr for the relationships between PA and Q and between PA and Wt are presented in Fig. 6. Similarly to the results in Fig. 4, r2 shows large variation with nLwthr, and also with k for the accumulated wind speed Wt. For the 645 nm band, the maximum r2 is reached at the same value of nLwthr(645) = 0.55 mW cm− 2 μm− 1 sr− 1 for the three relationships: PA645 vs. Vt (Fig. 4a), PA645 vs. Q (Fig. 6a), and PA645 vs. Wt (Fig. 6b). The k values for Vt and Wt, which yield the maximum correlation, are also about the same (0.5 and 0.55). This is not the case when the 555 nm band is

examined. The maximum r2 corresponds to nLwthr(555) that increases from 1.55 mW cm− 2 μm− 1 sr− 1 (k = 0.7) for the relationship involving Vt (Fig. 4b) to 2.15 mW cm− 2 μm− 1 sr− 1 (k = 0.5) for the relationship involving Wt (Fig. 6d). The differences in the patterns of variation in r2 for the two MODIS bands seen in Figs. 4 and 6 are not easily amenable to detailed explanation. Likely, these differences arise mostly from strong spectral variation in water-leaving radiance in response to variation in spectrally-dependent inherent optical properties (IOPs) of water. The water-leaving radiance at different wavelengths is also dependent to a different extent on the vertical structure of IOPs. The estimates of plume area calculated from each MODIS image in both bands using the optimum values of nLwthr based on the correlation analysis involving the parameters Vt, Q, and Wt are compared in Fig. 7. For the 645 nm band there is just one sets of data points illustrating PA645 because the optimum nLwthr(645) is the same for all three environmental parameters considered (see Table 2). For the 555 nm band the three sets of data points representing different estimates of PA555 correspond to different optimum values of nLwthr(555) associated with the three parameters (see Table 2). Although the compared estimates of PA differ to some extent, similar patterns of variability in the several estimates of PA shown in Fig. 7 are not surprising as the environmental variables covary to a large extent during rainstorm events (Fig. 3). Fig. 7 also shows that the plume areas calculated from the MODIS imagery at 645 nm are generally smaller than those calculated from imagery at 555 nm. In particular, PA645 is consistently smaller (by 8– 100%) than the PA555 estimates based on Vt and Q for all MODIS images examined. Only a small number of PA555 estimates based on Wt are somewhat smaller than the PA645 estimates (see the data for MODIS images 13–16 in Fig. 7). The differences between the bands are difficult to interpret in detail but not unexpected. The MODIS bands 1 (645 nm) and 4 (555 nm) have different spatial resolution of 250 m and 500 m at nadir, respectively. The higher resolution has naturally better potential for adequately resolving plume features, so the use of 250 m band is advantageous from that standpoint. In addition, the different wavelengths imply differences in the depth of the upper ocean layer “seen” by MODIS in these bands. Because of significantly higher water absorption in the red compared with green wavelengths,

Table 2 ̄ and Vt), the Tijuana River discharge (Q and Qt), the wind speed Results from the linear regression analysis between the plume area (PA645 and PA555 in km2) and the rainfall (PWS (W and Wt), the zonal and meridional wind stress (τEW, τEW,t, τNS, and τNS,t), and the tidal phase (TP). Parameter

r2

nLwthr

Linear equation

nLwthr and k Band 1 (645 nm)

Rainfall (mm) River flow (m3 s−1) Wind speed (m s−1) Zonal wind stress (kg m−1 s−2) Meridional wind stress (kg m−1 s−2)

Band 4 (555 nm)

Tidal phase (min) Rainfall (mm) River flow (m3 s−1) Wind speed (m s−1) Zonal wind stress (kg m−1 s−2) Meridional wind stress (kg m−1 s−2) Tidal phase (min)

0.63 0.77 0.56 0.11 0.10 0.83 0.08 0.89 0.07 0.19 0.45 0.62 0.75 0.53 0.36 0.09 0.79 0.12 0.84 0.06 0.21 0.43

1.00 (0.55; 0.55 (0.30; 0.45 (0.55; 0.30 (2.00; 1.85 (0.40; 0.30 3.00 (1.55; 1.85 (1.35; 1.85 (2.15; 1.55 (2.95; 0.60 (1.60; 1.60

0.5) 0.5) 0.55) 0.15) 0.55)

0.70) 0.05) 0.5) 0.05) 0.55)

̄ + 2.354 PA645 = 1.631 ⁎ PWS PA645 = 1.226 ⁎ Vt + 2.066 PA645 = 0.6318 ⁎ Q + 11.39 PA645 = 0.3437 ⁎ Qt + 38.92 PA645 = − 7.497 ⁎ W + 48.89 PA645 = 8.371 ⁎ Wt − 40.51 PA645 = − 917.9 ⁎ τEW + 51.64 PA645 = 80.43 ⁎ τEW,t − 0.4666 PA645 = − 112.4 ⁎ τNS + 1.744 PA645 = 747.4 ⁎ τNS,t + 20.33 PA645 = − 0.08651 ⁎ TP + 72.27 ̄ + 0.9309 PA555 = 1.455 ⁎ PWS PA555 = 1.236 ⁎ Vt + 10.22 PA555 = 0.6297 ⁎ Q + 17.53 PA555 = 0.9154 ⁎ Qt + 42.53 PA555 = − 6.792 ⁎ W + 47.14 PA555 = 8.029 ⁎ Wt − 34.00 PA555 = − 1123 ⁎ τEW + 48.53 PA555 = 645.7 ⁎ τEW,t − 1.513 PA555 = 2196 ⁎ τNS + 114.4 PA555 = 797.3 ⁎ τNS,t + 25.96 PA555 = − 0.08449 ⁎ TP + 65.24

The rainfall data represent average quantities obtained from measurements within the composite watershed WS7, which includes seven hydrologic units from the US part of Tijuana unit in the south to the San Dieguito unit in the north. For each environmental parameter (except for TP), the analysis was made for both the daily average value and the accumulated value (the latter denoted by subscript t). Results are presented for two MODIS bands and include: the coefficient of determination, r2, the optimum threshold value of normalized water-leaving radiance, nLwthr (in mW cm− 2 μm− 1 sr− 1), the parameter k, and the best fit linear equation.

F. Lahet, D. Stramski / Remote Sensing of Environment 114 (2010) 332–344

339

Fig. 6. The coefficient of determination, r2, of the linear relationship between the plume area, PA, and environmental factors as a function of the threshold value of satellite-derived normalized water-leaving radiance, nLwthr. Panels (a) and (c) are for the relationship between PA and the mean daily discharge of Tijuana River, Q. Panels (b) and (d) are for the relationship between PA and the accumulated value of wind speed, Wt. In this case, the correlation curves are plotted for various values of parameter k. The left-hand panels are for the MODIS band 1 centered at 645 nm and the right-hand panels are for the band 4 at 555 nm.

Fig. 7. The plume area, PA, calculated for each of the eighteen MODIS images considered in this study. The calculations were based on the optimum threshold values of normalized waterleaving radiance, nLwthr, given in Table 2. Specifically, the values of PA calculated from two MODIS bands are shown: for nLwthr(645) = 0.55 mW cm− 2 μm− 1 sr− 1 based on correlation analysis with accumulated precipitated water Vt (circles), mean daily discharge of Tijuana River Q (circles), and accumulated wind speed Wt (circles), and for nLwthr(555)= 1.55 mW cm−2 μm− 1 sr− 1 based on Vt (crosses), for nLwthr(555) = 1.85 mW cm− 2 μm− 1 sr− 1 based on Q (triangles), and for nLwthr(555)= 2.15 mW cm− 2 μm− 1 sr− 1 based on Wt (solid circles). The consecutive MODIS image numbers from 1 through 18 at the bottom horizontal axis correspond to dates of image acquisition (see Table 1). The time period of our study from December 28, 2004 through March 7, 2005 is indicated at the top axis.

340

F. Lahet, D. Stramski / Remote Sensing of Environment 114 (2010) 332–344

Fig. 8. MODIS images illustrating the spatial extent of turbid plumes observed during rainstorm events in the San Diego region. The images were acquired at the MODIS band 1 centered at 645 nm on the dates indicated in the upper right corner of the graphs. The study area is contained within the white box. Plume pixels are shown in white. Black pixels correspond to the lack of valid ocean color data. Land is shown in grey. The upper map on the left-hand side also shows the location of several geographic features and components within the region: Mission Bay (MB), San Diego River mouth (SDR), Point Loma (PL), San Diego Bay (SDBa), Tijuana River (TR), the US–Mexican border (B), and Los Buenos Creek mouth (LBC).

the surface layer seen within the 645 nm band is expected to be shallower than within the 555 nm band. Remote sensing at different wavebands is also differentially sensitive to variations in water optical properties associated with particulate and dissolved materials. Aerial multispectral imagery shows spectral reflectance differences between plumes that are freshly discharged and older parts of the plumes. Fresh plumes reflect more light at longer wavelengths than older plumes (Jan Svejkovsky, personal communication). This observation can likely be attributed to sinking of suspended sediments. This process is more intensive in the thin upper layer and, as such, might be easily resolved by remote sensing at longer wavelengths. 4.3. Characterization of plumes during selected rainstorm events In southern California, rainstorms are usually short episodic events occurring in winter. The turbid plumes associated with inputs of terrigenous and anthropogenic materials into the coastal ocean can persist for several days, and occasionally perhaps weeks (Nezlin et al., 2005; Warrick et al., 2007). Freshwater from rivers and streams quickly stratifies into a buoyant plume when it reaches the ocean, where the dispersal transport depends primarily on river plume inertia, wind, and current forcing (Washburn et al, 2003; Warrick et al., 2004a, Nezlin & DiGiacomo 2005; Warrick et al., 2007).

We selected six MODIS images to demonstrate characteristic plume features during two rainstorm events in the study area. Southerly winds force northward near-shore currents resulting in northward plume propagation; northerly winds result in plumes extended southward. The image of December 30, 2004 (Figs. 8a and 9a) illustrates the advection of the plume toward the north/northwest and the image of January 12, 2005 (Figs. 8b and 9b) towards the south. Both images were acquired during the same intense rainstorm event that lasted more than two weeks. The four remaining images (01/29/2005, 01/31/2005, 02/ 02/2005, and 02/03/2005, Figs. 8c–f and 9c–f) illustrate the progression from plume formation during a less intense storm event to advection towards the south and dispersion after the event. Fig. 8 shows the images at the 250 m spatial resolution obtained within the 645 nm waveband. Fig. 9 shows the 500 m resolution data within the 555 nm waveband. The plume is represented by white pixels with the normalized water-leaving radiance greater than the optimum threshold values determined from the relationships PA vs. Vt (see Table 2). The maps of surface currents for the days of image analysis are shown in Fig. 10. 4.3.1. MODIS images of December 30, 2004 and January 12, 2005 December 30, 2004 was the third day of the first intense rainstorm event that occurred during the investigated period (Table 1). From the

F. Lahet, D. Stramski / Remote Sensing of Environment 114 (2010) 332–344

341

Fig. 9. As in Fig. 8 but for the MODIS band 4 centered at 555 nm.

beginning of this event on December 28, 2004 until its end on January 12, 2005, it rained almost every day in the San Diego and Tijuana ̄ watersheds. On December 30, PWS was about 3.3 mm and Vt about 55 mm. On December 28 and December 29, heavy rains with daily precipitation of 19 mm and 45 mm respectively, and wind speeds >6 m s− 1 were observed (Fig. 3). The winds were blowing from the southeast on December 28 and then from the southwest on December 29 and 30. The average wind speed decreased to ∼1.8 m s− 1 on December 30. Q was very low before the storm (<0.4 m3 s− 1 since December 15), increased markedly on December 29 (4.1 m3 s− 1), and reached a maximum of about 100 m3 s− 1 on December 30 (Fig. 3). The latter value is the highest observed during the investigated time period. Current data within the study area on December 30 indicates that the direction was generally northwest and magnitude was ∼25 cm s− 1 (Fig. 10a). MODIS images acquired on December 30 suggest that the plume was advected north to northwest (Figs. 8a and 9a). The plume area is 60– 76 km2 and the offshore extent exceeds 12 km at the S1 section located near Point Loma entrance to the San Diego Bay (Table 3, Fig. 1). The spatial structure of the plume appears to reflect sources of turbid water discharge from Tijuana River (TR), San Diego Bay (SDBa), and San Diego River (SDR) north of the study area (see Fig. 8a). This structure is most likely dominated by the combination of inertia-related effects and northward current forcing. Resuspension processes in areas with relatively shallow bathymetry may have also contributed to plume turbidity given that the winds exceeded 6 ms− 1 on preceding days. The study on the San Pedro

Shelf (Los Angeles area) showed that bottom sediments are effectively resuspended by waves and currents at bottom depths less than about 15 m (Washburn et al., 1992). In our study area such shallow bathymetry occurs generally within a narrow zone along the shore. Somewhat larger offshore extent of shallow depths occurs off the entrance to San Diego Bay and off the Tijuana River mouth (see Fig. 1). January 12, 2005 was the last day of the rainstorm event that started over two weeks earlier. Before the acquisition of the MODIS ̄ image on January 12, PWS ranged from 9 to 16 mm from January 7 through January 10, and reached 30 mm on January 11 (Fig. 3). This explains high values of Vt of about 50 mm on January 12. There was a ̄ of 4.12 mm. Q was less than 8 m3 s− 1 light rain on January 12 with PWS during the preceding week but reached 49 m3 s− 1 on January 12. Winds between 3.5 and 6.7 m s− 1 were blowing from the southeast from January 7 through January 10. The average wind speed increased on January 11 (∼7.3 m s− 1) and the direction changed to southwest. On January 12 the wind direction changed to northwest and the speed decreased to ∼2.5 m s− 1. At the northern end of the study area, currents of about 30 cm s− 1 were flowing southwest (Fig. 10b). In the southern part, the direction of currents varied generally with the distance from coastline from southward to southeastward, and to southwestward. The magnitude of currents generally decreased towards the south. Consistent with the general pattern of currents and winds are the MODIS images of January 12, which show an advection of the plume towards the south (Figs. 8b and 9b). Compared to December 30, the plume on January 12 extends significantly further south, well beyond the

342

F. Lahet, D. Stramski / Remote Sensing of Environment 114 (2010) 332–344

Fig. 10. Mean daily surface currents obtained from measurements with HF radars located in four different locations within the region (see text for details, the figures provided by the San Diego Coastal Ocean Observing System). The current fields are shown for the same dates as those in Figs. 8 and 9.

study area. The plume is large with an area comparable to that observed two weeks before (Table 3). The offshore extent of the plume is similar (∼7–8 km) throughout much of the study area. Table 3 Area (PA) and offshore extent (S1, S2, S3, and S4) of the plume calculated from MODIS images at bands 1 and 4 within the study area.

Band 1 (645 nm)

Band 4 (555 nm)

12/30/04 01/12/05 01/29/05 01/31/05 02/02/05 02/03/05 12/30/04 01/12/05 01/29/05 01/31/05 02/02/05 02/03/05

PA (km2)

S1 (km)

S2 (km)

S3 (km)

S4 (km)

60.2 57.0 20.4 16.7 3.29 0.56 76.0 89.6 56.3 37.2 20.8 2.96

12.75 8.75 0 2.50 0 ND 12.50 7.50 6.50 6.50 5.00 ND

5.75 5.50 3.50 ND 0 ND 6.50 7.50 5.00 4.50 1.50 ND

3.50 8.50 4.00 3.00 1.75 ND 3.50 8.00 5.00 3.50 2.50 ND

1.50 6.75 ND 1.50 ND ND 1.50 7.00 3.50 3.50 2.00 ND

The plume area was calculated from the optimum threshold values of nLwthr and parameter k corresponding to the relationships PA vs. Vt (see Table 2). The offshore extent of the plume was estimated across four sections (see also Fig. 1): S1 located at Point Loma latitude (32°40′N, 117°10′W), S2 between Point Loma and Tijuana River mouth (32°37′N, 117°08′W), S3 at Tijuana River mouth (32°33′N, 117°08′W), and S4 between Tijuana River mouth and the southern end of the study area (32°29′N, 117°07′W). ND indicates no valid satellite data.

4.3.2. Sequence of MODIS images from January 29, 2005 through February 3, 2005 The second rainy episode during the investigated time period occurred at the end of January 2005 and lasted 5 days (see Table 3). Compared to the previous rainstorm, this event was less intense with daily precipitation in the range from 0.3 to 6.7 mm, river flow from about zero to 0.74 m3 s− 1, and winds up to about 4 m s− 1 with variable direction alternating between northwest and southwest (Fig. 3). On January 29, the last day of the event, P̄WS was about 4 mm, Vt was in the range 8–9 mm, and the wind was blowing from the northwest. The currents of about 13 cm s− 1 were oriented southeast (Fig. 10c). There was still a very weak rain on January 31 (P̄WS = 0.28 mm) but no rain on the following days. Vt showed a significant decrease after January 29. Weak northwest winds (∼2 m s− 1) were observed from January 30 through February 1. On February 2 and 3, the winds increased to ∼4– 5 m s− 1 and changed direction to northeast. The current data show predominantly southward flow during that period (Fig. 10d–f). The sequence of four MODIS images for the period of January 29 through February 3 illustrates how the plume was advected towards the south after its formation during a moderate rain event, and eventually dispersed to the point below the defining threshold on February 3 (Figs. 8c–f and 9c–f). Under very low or no precipitation and low discharge from Tijuana River, the evolution of the plume was linked primarily to the direction and strength of the wind and alongshore

F. Lahet, D. Stramski / Remote Sensing of Environment 114 (2010) 332–344

currents. The images from January 29 through February 2 clearly show that the plume elongates along the north–south direction and is being transported towards the south (Figs. 8c–e and 10c–e). On February 3 only a small number of plume pixels were identified. This result can be attributed to the main mechanisms of plume disappearance, which include dispersion and dilution with ambient waters, and settling of suspended particulate matter. For the images showing a sizeable plume, the plume area is significantly smaller in the 645 nm band compared with the 555 nm band, which is consistent with the data from the previous rainstorm (Table 3). The image in the 645 nm band appears to better reveal the discharge sources of turbid waters than the 555 nm band. On January 29 the plume seen in the 645 nm band is broken up into three parts, which are not as readily observed in the 555 nm band. The three parts seen in the 645 nm band appear to be associated with the San Diego Bay (small plume near Point Loma), Tijuana River (large plume in the central part of the image), and Los Buenos Creek and the city of Rosarito (plume features south of the study area where the image is partly obscured by significant cloud cover). This latter plume is actually well seen in the next image of January 31. 5. Conclusions High correlation between the area of turbid plumes estimated from MODIS satellite imagery in the San Diego coastal region and several environmental factors, such as rainfall, river flow, and wind speed, supports earlier studies showing that optical remote sensing can serve as a means for monitoring a discharge of terrigenous and anthropogenic particulate and dissolved materials and subsequent evolution of turbid plumes in the coastal ocean during rainstorm events. The remote sensing of such plumes is important because stormwater runoff is a primary source of contaminants and water quality degradation in many coastal regions, with potentially negative impacts on humans and ecosystems (e.g., Noble et al., 2003; Schiff & Bay, 2003). In southern California, high pressure systems often follow winter storms, which typically serve to clear the skies. Such conditions mitigate to some extent the potential problem of cloud cover that normally prevents the acquisition of ocean color imagery (Nezlin et al., 2007). By using the MODIS imagery at a spatial resolution of 250 m and 500 m, our study extends the previous utilization of 1 km satellite data from SeaWiFS in similar studies of southern California coastal waters (Nezlin et al., 2005). We used a correlation analysis between the plume area and the amount of precipitated water accumulated during a storm event over the San Diego and Tijuana watersheds to determine the threshold water-leaving radiance for distinguishing the plume from ambient ocean waters. This method limits arbitrariness in determining the threshold criteria for defining the plume boundaries and may be broadly applicable beyond our study area. However, we caution against indiscriminate use of our quantitative results of threshold radiance and empirical relationships between the plume area and rainfall in other regions. This is because these results will depend on various environmental conditions such as land-use characteristics, watershed size, land topography, soil conditions, etc. The analysis of MODIS images in the study area, which extends along the San Diego coastline from approximately Point Loma in the north to 10 km south of the US–Mexican border, demonstrated characteristic features of plume dynamics under various environmental conditions, such as advection towards the south or north during an intense rainstorm event, and the southward advection accompanied by dispersion within several days after the storm. Under heavy rainfall the maximum offshore extent of the plume exceeded 12 km and the plume was continuous for tens of kilometers along the shore. Major sources of stormwater runoff along the coastline were discernible on the images, in particular the Tijuana River discharge and the outflow of turbid waters into the coastal ocean from San Diego Bay. The plume area estimated from MODIS band 4 at 555 nm was generally larger than that calculated

343

from MODIS band 1 at 645 nm. These differences can be attributed largely to differences in the inherent optical properties of seawater in the green and red wavebands, which result in differences in the depth of the upper ocean layer “seen” in these bands. The differences in the spatial resolution (250 m for band 1, 500 m for band 4) also play a role. The use of both bands offers advantages. Whereas the 645 nm band provides information with higher spatial resolution and appears to better reveal the location of distinctive sources of runoff along the coastline, the remotely-sensed information at the 555 nm band originates within a thicker layer of surface water compared with 645 nm band. Acknowledgements This study was supported by the National Aeronautics and Space Administration (Earth Observing System Interdisciplinary Science Program, NASA EOS/IDS Grant NNG04GK50G awarded to Dariusz Stramski). We would like to thank the NASA Goddard Space Flight Center for the production of MODIS images. We thank Eric Terrill and Mark Otero from the San Diego Coastal Ocean Observing System at Scripps Institution of Oceanography for river discharge and surface current data. Precipitation data were provided by Rand Allan from the County of San Diego, Department of Public Works, Raúl Larios, Alejandro González, Javier Espinosa from the Comisión Nacional del Agua in Mexico, the National Oceanic and Atmospheric Administration National Climatic Data Center Climate Data Online (NCDC/NNDC CDO), and the United States Geological Survey (USGS). We also thank Nikolay Nezlin from Southern California Coastal Water Research Project, Jan Svejkovsky from Ocean Imaging Corp., Rick Reynolds from Scripps Institution of Oceanography, and two anonymous reviewers for their invaluable comments on the manuscript. References Ackerman, D., & Weisberg, S. B. (2003). Relationship between rainfall and beach bacterial concentrations on Santa Monica Bay beaches. Journal of Water and Health, 1, 85−89. Ahn, J. H., Grant, S. B., Surbeck, C. Q., DiGiacomo, P. M., Nezlin, N. P., & Jiang, S. (2005). Coastal water quality impact of stormwater runoff from an urban watershed in southern California. Environmental Science and Technology, 39, 5940−5953. Bray, N. A., Keyes, A., & Morawitz, W. M. L. (1999). The California current system in the Southern California Bight and the Santa Barbara Channel. Journal of Geophysical Research, 104, 7695−7714. Chadwick, D. B., & Largier, J. L. (1999). Tidal exchange at the bay–ocean boundary. Journal of Geophysical Research, 104, 29901−29924. Chao, S. Y. (1998). Hyperpycnal and buoyant plumes from a sediment-laden river. Journal of Geophysical Research, 103, 3067−3081. Characklis, G. W., & Wiesner, M. R. (1997). Particles, metals, and water quality in runoff from large urban watershed. Journal of Environmental Engineering, 753−759. Chen, Z., Hu, C., & Muller-Karger, F. (2007). Monitoring turbidity in Tampa Bay using MODIS/Aqua 250-m imagery. Remote Sensing of Environment, 109, 207−220. Davis, P., Shokouhian, M., & Ni, S. (2001). Loading estimates of lead, copper, cadmium, and zinc in urban runoff from specific sources. Chemosphere, 44, 997−1009. Dinnel, S. P., Schroeder, W. W., & Wiseman, W. J. (1990). Estuarine-shelf exchange using Landsat images of discharge plumes. Journal of Coastal Research, 6, 789−799. Garvine, R. W. (1995). A dynamical system for classifying buoyant coastal discharges. Continental Shelf Research, 15, 1585−1596. Hu, C., Chen, Z., Clayton, T., Swarnzenski, P., Brock, J., & Muller-Karger, F. (2004). Assessment of estuarine water-quality indicators using MODIS medium-resolution bands: Initial results from Tampa Bay, Florida. Remote Sensing of Environment, 93, 423−441. Inman, D. L., & Jenkins, S. A. (1999). Climate change and the episodicity of sediment flux of small California rivers. Journal of Geology, 107, 251−270. Isla, N. M., & Lee, J. L. (2006). Climate of San Diego, California.: National Oceanic and Atmospheric Administration, National Weather Service Technical Memorandum NWS WR-275. Kim, S. Y., Terrill, E., & Cornuelle, B. (2007). Objectively mapping HF radar-derived surface current data using measured and idealized covariance matrices. Journal of Geophysical Research, 112, C06021. doi:10.1029/2006JC003756. Large, W. G., & Pond, S. (1981). Open ocean momentum flux measurements in moderate to strong winds. Journal of Physical Oceanography, 11, 324−336. Lentz, S. J., & Winant, C. D. (1986). Subinertial currents on the Southern California Shelf. Journal of Physical Oceanography, 16, 1737−1750. Nezlin, N. P., & DiGiacomo, P. M. (2005). Satellite ocean color observations of stormwater runoff plumes along the San Pedro Shelf (southern California) during 1997–2003. Continental Shelf Research, 25, 1692−1711.

344

F. Lahet, D. Stramski / Remote Sensing of Environment 114 (2010) 332–344

Nezlin, N. P., DiGiacomo, P. M., Diehl, D. W., Jones, B. H., Johnson, S. C., Mengel, M. J., et al. (2008). Stormwater plume detection by MODIS imagery in the southern California coastal ocean. Estuarine Coastal and Shelf Science, 80, 141−152. Nezlin, N. P., DiGiacomo, P., Stein, E. D., & Ackerman, D. (2005). Stormwater runoff plumes observed by SeaWiFS radiometer in the southern California Bight. Remote Sensing of Environment, 98, 494−510. Nezlin, N. P., & Stein, E. D. (2005). Spatial and temporal patterns of remotely-sensed and field-measured rainfall in southern California. Remote Sensing of Environment, 96, 228−245. Nezlin, N. P., Weisberg, S. B., & Diehl, D. W. (2007). Relative availability of satellite imagery and ship-based sampling for assessment of stormwater runoff plumes in coastal southern California. Estuarine, Coastal and Shelf Science, 71, 250−258. NOAA National Weather Service. (2007). A history of significant weather events in Southern California. San Diego. Noble, R. T., Weisberg, S. B., Leecaster, M. K., McGee, C. D., Dorsey, J. H., Vainik, P., et al. (2003). Storm effects on regional beach water quality along the southern California shoreline. Journal of Water and Health, 1, 23−31. Raphael, M. N. (2003). The Santa Ana winds of California. Earth Interactions, 7, 1−13. Roughan, M., Terrill, E. J., Largier, J. L., & Otero, M. P. (2005). Observations of divergence and upwelling around Point Loma, California. Journal of Geophysical Research, 110. doi:10.1029/2004JC002662. Schiff, K. C., Allen, M. J., Zeng, E. Y., & Bay, S. M. (2000). Southern California. Marine Pollution Bulletin, 41, 76−93. Schiff, K. C., & Bay, S. (2003). Impacts of stormwater discharges on the nearshore benthic environment of Santa Monica Bay. Marine Environmental Research, 56, 225−243. Shutler, J. D., Land, P. E., Smyth, T. J., & Groom, S. B. (2007). Extending the MODIS 1 km ocean colour atmospheric correction to the MODIS 500 m bands and 500 m chlorophyll-a estimation towards coastal and estuarine monitoring. Remote Sensing of Environment, 107, 521−532. Stumpf, R. P., Gelfenbaum, G., & Pennock, J. R. (1993). Wind and tidal forcing of a buoyant plume, Mobile Bay, Alabama. Continental Shelf Research, 13, 1281−1301. Tran, K., Yu, C. C., & Zeng, E. Y. (1997). Organic pollutants in the coastal environment off San Diego, California. 2. Petrogenic and biogenic sources of aliphatic hydrocarbons. Environmental Toxicology and Chemistry, 16, 189−195.

Wang, M. (2007). Remote sensing of the ocean contributions from ultraviolet to nearinfrared using the shortwave infrared bands: Simulations. Applied Optics, 46, 1535−1547. Wang, M., & Shi, W. (2005). Estimation of ocean contribution at the MODIS nearinfrared wavelengths along the east coast of the U.S.: Two case studies. Geophysical Research Letters, 32, L13606. doi:10.1029/2005GL022917. Warrick, J. A., Mertes, L. A. K., Siegel, D. A., & Washburn, L. (2004). Dispersal forcing of a southern California river plume, based on field and remote sensing observations. Geo-Marine Letters, 24, 46−52. Warrick, J. A., Mertes, L. A. K., Washburn, L., & Siegel, D. A. (2004). A conceptual model for river water and sediment dispersal in the Santa Barbara Channel, California. Continental Shelf Research, 24, 2029−2043. Warrick, J. A., DiGiacomo, P. M., Weisberg, S. B., Nezlin, N. P., Mengel, M., Jones, B. H., et al. (2007). River plume patterns and dynamics within the Southern California Bight. Continental Shelf Research, 27, 2427−2448. Washburn, L., Jones, B. J., Bratkowich, A. W., Dickey, T. D., & Chen, M. S. (1992). Mixing, dispersion, and resuspension in vicinity of ocean wastewater plume. Journal of Hydraulic Engineering, 118, 38−58. Washburn, L., McClure, K. A., Jones, B., & Bay, S. M. (2003). Spatial scales and evolution of stormwater plumes in Santa Monica Bay. Marine Environmental Research, 56, 103−125. Winant, C. D., & Bratkovich, A. W. (1981). Temperature and currents on the Southern California shelf: A description of the variability. Journal of Physical Oceanography, 11, 71−86. Zeng, E. Y., Khan, A. R., & Tran, K. (1997). Organic pollutants in the coastal marine environment off San Diego, California. 3. Using linear alkylbenzenes to trace sewagederived organic materials. Environmental Toxicology and Chemistry, 16, 196−201. Zeng, E. Y., & Vista, C. L. (1997). Organic pollutants in the coastal environment off San Diego, California. 1. Source identification and assessment by compositional indices of polycyclic aromatic hydrocarbons. Environmental Toxicology and Chemistry, 16, 179−188.